nasa-ibm-ai4science/surya-bench-ar-segmentation

---

A Dataset of Binary Maps of Active Regions with Polarity Inversion Lines

Dataset Summary

This dataset provides hourly binary segmentation maps (4096×4096 resolution) derived from Solar Dynamics Observatory (SDO) / Helioseismic and Magnetic Imager (HMI) line-of-sight magnetograms. The maps highlight regions containing Active Regions (ARs) and Polarity Inversion Lines (PILs). The dataset spans observations from May 13, 2010 to December 31, 2024 and is intended for image segmentation tasks using full-disk 4096×4096 magnetogram rasters as input.

Supported Tasks and Applications

• segmentation: Binary segmentation of solar magnetograms to identify ARs with strong PILs.

Dataset Structure

Data Files

• train.csv: Instances from Feb 15 to Dec 31 of each year between 2010–2019, day index [45:]
• validation.csv: Instances from Jan 15 to Jan 31 of each year between 2010–2019, day index [14:31]
• test.csv: All instances from each year between 2020–2024, day index [:]
• leaky_validation.csv: Instances from Jan 1 to 14 and Feb 1-14 of each year between 2010–2019, day indices [0:14] and [31:45]

Features

Each .h5 file contains two keys:

• intersection: Binary map of polarity inversion lines (PILs)
• union_with_intersect: Binary map of active regions that contain PILs

Dataset Details

Field	Description
Temporal Coverage	May 13, 2010 – Dec 31, 2024
Data Format	HDF5 (.h5)
Data Shape	[2, 4096, 4096] per file
Total Instances	Total 128,352 instances
Cadence	1 hour
Size per File	~ 53kB
Total Dataset Size	~ 1.2GB

Authors

• Jinsu Hong, jhong36@gsu.edu
• Kang Yang, kyang30@gsu.edu
• Berkay Aydin, baydin2@gsu.edu